Think-on-Graph: Deep and Responsible Reasoning of LLMs on Knowledge Graphs
LLMs perform step-by-step reasoning by traversing knowledge graph paths.
Think-on-Graph (ToG) introduces a paradigm where LLMs perform reasoning by actively traversing knowledge graph paths. The LLM iteratively explores the graph, deciding which edges to follow based on the reasoning task. Achieves state-of-the-art on multi-hop QA benchmarks with interpretable reasoning chains.
Tags
reasoningknowledge-graphllmmulti-hopthink-on-graph